Anthropic’s Head of Platform Engineering attempts to offer reassurance about AI sandbox escapes:
1) It only happens in training ‼️
She also claims that in production, some of the sandbox incidents "could never really happen."
2) It's happening at every AI lab ‼️
"Long-running agents is a relatively nascent thing that everybody is building and trying to deal with... Some of that is just happening as we go."
-
@katelyn_lesse on
@latentspacepod